Real-Time Railway Track Monitoring Using Yolov8 and Embedded Systems for Automated Defect Detection

Abstract

The Railway track monitoring is essential for ensuring the safety, reliability, and longevity of rail transportation infrastructure. Traditional inspection methods such as manual surveys and track circuit-based systems are labour intensive, time consuming, and prone to human error, which results in delayed fault detection and increased maintenance risks. Addressing these challenges, this study presents the Railway Track Monitoring Tool (RT-MT), an intelligent real time defect detection system that integrates machine learning with embedded electronics to enhance railway infrastructure monitoring. The RT-MT employs a Raspberry Pi 5 (4GB), a Universal Serial Bus (USB) 1080p camera, a SIM800C GSM module, and a UBLOX NEO 6M Global Positioning System (GPS) module to automate track inspections. The camera captures railway track images which are processed using a You Only Look Once (YOLO) v8 based object detection model deployed on the Raspberry Pi to identify cracks and missing fasteners, which are critical defects affecting track integrity. A dataset consisting of 1,000 crack images and 500 fastener images was used to train the model, ensuring reliable defect classification. When a defect is detected, GPS coordinates are extracted and an alert is sent through a messaging system to maintenance teams for immediate action. A spam prevention mechanism prevents redundant alerts by triggering notifications only when a defect is detected beyond a 50-meter radius from the previous detection. A web-based monitoring system provides real time defect tracking, visualization, and data logging for data driven maintenance decisions. The Railway Track Monitoring Tool, validated on a prototype railway track, offers a cost effective, scalable, and automated solution, addressing the limitations of traditional methods while improving railway safety and operational efficiency.

Introduction

1. Overview and Importance

Railway track monitoring is critical for ensuring safety, operational efficiency, and infrastructure longevity. Traditional manual inspections are labor-intensive, slow, and prone to human error. The growing demands of high-speed and expansive rail networks have driven the need for automated, real-time monitoring systems.

2. Technological Advancements

Recent innovations leverage:

• Artificial Intelligence (AI) and Machine Learning (ML) for object detection.

• Internet of Things (IoT) for real-time alerts and geolocation.

• Embedded systems like Raspberry Pi, cameras, GPS, and GSM modules.

These technologies enable proactive, predictive maintenance, reduce costs, and improve safety by continuously monitoring track conditions and sending instant alerts when defects are detected.

3. RT-MT System Design

The Real-Time Monitoring Tool (RT-MT) is an automated inspection system featuring:

• A mobile rover with a USB camera, GPS, GSM, night-vision, and infrared sensors.

• Deep learning (DL) using YOLOv8 to detect defects like cracks and missing fasteners.

• Real-time alerts via SMS with location and severity details.

4. Key Contributions

• Implementation of YOLOv8 for real-time railway defect detection.

• Use of Raspberry Pi 5, camera, GPS, and GSM for embedded monitoring and alerts.

• A web-based monitoring interface for real-time visualization and data-driven maintenance.

5. Literature Review Insights

• Wang et al.: Used YOLOv8n for defect detection; achieved 94%+ accuracy but lacked edge computing optimization.

• Lin et al.: Used YOLOv3 for fastener inspection; effective but lacked an integrated end-to-end system.

• Sun et al.: Used CNNs with acceleration data for joint detection; missed surface collapse and batter detection.

• Chellaswamy et al.: Used accelerometers and mathematical modeling; limited by GSM signal strength.

• Bogacz et al.: Modeled dynamic track behavior using periodic sleeper spacing; didn’t address real-world defects.

• Anand et al.: Compared NoSQL databases; analysis lacked direct performance benchmarks.

6. Materials and Methods

• Image/Video Capture: Real-time frames of railway tracks are captured.

• Preprocessing: Includes noise reduction, resizing, normalization, and augmentation.

• Detection: YOLOv8 detects anomalies; GPS records location; GSM sends alerts.

• Dataset: 1,500 labeled images (Cracks and Fasteners); 80% for training, 10% validation, 10% testing.

• Labeling: Manual annotation using Roboflow with bounding boxes and severity classification.

7. Model Architecture

• YOLOv8 Structure: Consists of backbone (CNN for feature extraction), neck (FPN, PAN), and head (for bounding boxes and class labels).

• Ensures multi-scale accuracy and real-time processing.

8. Prototype Development

• Designed using Onshape for scalable rover and track models.

• 3D printing used to fabricate fasteners with artificial defects.

• Real-world fabrication involved metal I-beams and welded rails for realistic track conditions.

• Rover integrated with sensors and tested under diverse lighting and environmental conditions.

9. Implementation and Testing

• Hardware: Raspberry Pi 5, USB Camera, GPS, GSM.

• Real-world testing confirmed:

  • Accurate and timely detection of defects.

  • Reliable alert system with geolocation.

  • Strong performance in varied light and weather conditions.

Conclusion

The Railway Track Monitoring Tool integrates machine learning-based object detection with embedded hardware to develop an efficient real-time defect detection system. By focusing on cracks and missing fasteners, the system ensures precise and reliable railway infrastructure monitoring. The custom-welded prototype, designed using 3D modelling software, provided a controlled environment to evaluate defect detection and classification capabilities. The USB 1080p camera, integrated with Raspberry Pi 5, enabled continuous image acquisition and processing, while GPS-based fault mapping ensured accurate localization of detected defects. The SIM800C GSM module successfully transmitted fault notifications, allowing maintenance teams to respond promptly. The spam prevention mechanism minimized redundant alerts within a 50-meter radius, reducing unnecessary notifications and enhancing operational efficiency. By integrating AI-driven object detection with embedded hardware components including Raspberry Pi, GSM, GPS, and camera modules, this research presents a scalable and cost-effective solution for railway track monitoring. The prototype demonstrates the feasibility of real-time defect detection and automated alerting, contributing to enhanced railway safety and predictive maintenance. Future improvements include higher-resolution imaging, advanced deep learning models, and large-scale deployment on operational railway networks. The combination of deep learning, embedded electronics, and real-time monitoring establishes an innovative approach to railway infrastructure assessment.

References

[1] Bianchi, G., Fanelli, C., Freddi, F., Giuliani, F., & La Placa, A. (2025). Systematic review railway infrastructure monitoring: From classic techniques to predictive maintenance. Advances in Mechanical Engineering, 17(1), 16878132241285631. [2] Banic, M., Ristic-Durrant, D., Madic, M., Klapper, A., Trifunovic, M., Simonovic, M., & Fischer, S. (2025). The Use of Earth Observation Data for Railway Infrastructure Monitoring—A Review. Infrastructures, 10(3), 66. [3] Kurhan, D., Kovalchuk, V., Markul, R., &Kovalskyi, D. (2025). Development of Devices for Long-Term Railway Track Condition Monitoring: Review of Sensor Varieties. Acta PolytechnicaHungarica, 22(4). [4] Winarno, A., Sandhy, R. I. D., Nazmi, N., Priatomo, H. R., Suwastono, A., Charisma, G. E., & Bahiuddin, I. (2025). Roles of Vibration-Based Machine Learning Algorithms in Railway Vehicle Monitoring for Track Condition Assessment: A Review. Journal of Vibration Engineering & Technologies, 13(4), 242. [5] Pereira, P., Ribeiro, F., Motta, R., Neto, A. G., Bernucci, L., Júnior, E. F., ... & Melo, I. (2025). Perspectives of assessing the geometric condition of railway tracks using a device for measuring the topographic profile. Measurement, 249, 117020. [6] MajidiParast, S., Monemi, R. N., & Gelareh, S. (2025). A graph convolutional network for optimal intelligent predictive maintenance of railway tracks. Decision Analytics Journal, 14, 100542. [7] Mohammadi, M., Mosleh, A., Vale, C., Ribeiro, D., Montenegro, P., & Meixedo, A. (2025). Smart railways: AI-based track-side monitoring for wheel flat identification. Proceedings of the Institution of Mechanical Engineers, Part F: Journal of Rail and Rapid Transit, 09544097251313570. [8] Wang, Y., Zhang, K., Wang, L., & Wu, L. (2024). An improved YOLOv8 algorithm for rail surface defect detection. IEEE Access. [9] Lin, Y. W., Hsieh, C. C., Huang, W. H., Hsieh, S. L., & Hung, W. H. (2019, October). Railway track fasteners fault detection using deep learning. In 2019 IEEE Eurasia Conference on IOT, Communication and Engineering (ECICE) (pp. 187-190). IEEE. [10] Sun, Y., Liu, Y., & Yang, C. (2019, August). Railway joint detection using deep convolutional neural networks. In 2019 IEEE 15th International Conference on Automation Science and Engineering (CASE) (pp. 235-240). IEEE. [11] Chellaswamy, C., Duraichamy, S., Glaretsubin, P., & Vanathi, A. (2017, February). Optimized vehicle acceleration measurement for rail track condition monitoring. In 2017 2nd International Conference on Computing and Communications Technologies (ICCCT) (pp. 155-160). IEEE. [12] Bogacz, R., Czyczu?a, W., &Konowrocki, R. (2014). Influence of Sleepers Shape and Configuration on Track?Train Dynamics. Shock and Vibration, 2014(1), 393867. [13] Anand, V., & Rao, C. M. (2016, February). MongoDB and Oracle NoSQL: A technical critique for design decisions. In 2016 International Conference on Emerging Trends in Engineering, Technology and Science (ICETETS) (pp. 1-4). IEEE. [14] Yang, L., Sun, Y., Zhai, Z., & Wei, L. (2025, January). Evaluation of YOLOv8-Based Detection Techniques for Railway Track Anomalies Across Multiple Color Models. In 2025 IEEE 5th International Conference on Power, Electronics and Computer Applications (ICPECA) (pp. 791-795). IEEE. [15] Bai, T., Fu, H., Xu, G., Xin, X., Duan, J., & Zong, H. (2025). Crack-YOLOv8: Multi-class Crack Detection Method for Track Slabs Based on Improved YOLOv8. Measurement Science and Technology. [16] Kim, E. Y. (2025). Neues verkehrswissenschaftliches Journal NVJ-Ausgabe 35: 3D-Printed Scale Model for Detection of Railway Wheel Flats using Augmented Vibration Data from Axle Box. BoD–Books on Demand. [17] Aldao, E., Fernández-Pardo, L., Veiga-López, F., González-deSantos, L. M., & González-Jorge, H. (2025). Synthetic Data Generation Techniques for Enhancing Crack Detection in Railway Concrete Sleepers. Journal of Computing in Civil Engineering, 39(3), 04025032. [18] Qiu, S., Zaheer, Q., Hassan Shah, S. M. A., Shah, S. F. H., Wang, W., Ai, C., & Wang, J. (2025). LiDAR-Simulated Multimodal and Self-Supervised Contrastive Digital Twin Approach for Probabilistic Point Cloud Generation of Rail Fasteners. Journal of Computing in Civil Engineering, 39(2), 04025001. [19] Mian Qaisar, S., Say, D., Krichen, M., & Zidi, S. Artificial Intelligence Assistive Non-Destructive Testing of Welding Joints: A Review. Artificial Intelligence Assistive Non-Destructive Testing of Welding Joints: A Review. [20] Santamato, G., Tozzetti, L., Solazzi, M., Fedeli, E., & Di Pasquale, F. (2025). SmartRail: a system for the continuous monitoring of the track geometry based on embedded arrays of fiber optic sensors. IEEE Transactions on Intelligent Transportation Systems.

Copyright

Copyright © 2025 Nikhil Jones, Jovin Sebastian, Mohammed Mishal P S, Dr. Nelwin Raj N.R, Dr. Renjith R J. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.